Thomas Risse
Democratic Peace - Warlike Democracies?
A Social Constructivist Interpretation
of the Liberal Argument*
Introduction
The liberal argument that democratic political structures form a precondition
for stable peace orders in international relations has become conventional wisdom among Western policy-makers. Immanuel Kant’s postulate developed in
his ‘Perpetual Peace’ (1795[1991]) has been empirically substantiated. Peace
and conflict research has reached a consensus that democracies rarely fight
each other (Russett, 1993; Chan, 1993; Bueno de Mesquita et al., 1991). However, the ‘democratic peace’ only forms one part of the empirical finding.
Democracies are Janus-faced. While they do not fight each other, they are frequently involved in militarised disputes and war with authoritarian regimes.
Democratic peace despite warlike democracies?
This article does not pretend to add new data to the debate on democracy
and peace. It is about theory building. I argue that the two empirical findings on
the dyadic level concerning the war involvement of democracies are under-theorised. Most liberal theories of international relations assume that democracies
are inherently peaceful, while authoritarian regimes are considered intrinsically aggressive. However, these attempts at theorising about democracy and warinvolvement do not capture the different behaviour by democratic states
depending with which they are dealing. There is little empirical support for the
proposition that war-involvement of democracies mostly results from the need
to defend themselves against aggressive dictatorship. How is it then to be
explained that, on the one hand, democracies rarely fight each other and build
stable peace orders among themselves, but, on the other hand, can be rather
belligerent in their interactions with authoritarian regimes? Following ErnstOtto Czempiel (1986), Michael Doyle (1983, 1986), and Bruce Russett (1993), I
start from the so-called ‘normative explanation’ of the democratic peace. I
argue that this explanation can be considerably refined from a social-constructivist perspective. I claim that democracies to a large degree create their enemies and their friends — ’them’ and ‘us’ — by inferring either aggressive or de-
19
Thomas Risse
fensive motives from the domestic structures of their counterparts. Therefore,
they follow behavioural norms externalising their internal compromise-oriented and non-violent decision rules in their interactions with other democracies.
‘Pacific federations’ (Kant, 1795[1991]) or pluralistic security communities
(Deutsch, 1957) such as the European Union or the transatlantic relationship
can be easily created among democratic systems. The presumption of potential
enmity, however, creates a realist world of anarchy when democratic states
interact with authoritarian regimes. Limited ‘co-operation under anarchy’ is still
possible, but problems of defection and relative gains might hamper co-operative arrangements between democracies and non-democracies.
The article consists of three parts. First, I briefly review the empirical state of
the art. Second, I discuss the prevailing theoretical explanations for the dual
finding of ‘democratic peace’ and ‘warlike democracies’. I then present a
social-constructivist modification of the normative argument. I conclude with
suggestions for a research agenda to evaluate the argument presented in this
article.
The Empirical Findings: Democratic Peace - Warlike
Democracies?
The empirical findings on the correlation between domestic political structures
and foreign policy behaviour can be summarised in two statements:
1. Democracies are involved in as many wars as other types of political systems.
2. Democracies rarely fight each other.
Warlike Democracies?
Between 1816 (i.e., after the end of the Napoleonic wars) and 1976, democratic political systems were involved in as many militarised disputes, international
crises, and wars as authoritarian regimes and proportionate to their percentage
among states (Maoz and Abdolali, 1989; see also Small and Singer, 1976; Gleditsch and Hegre, 1995). The result does not change significantly when only
‘politically relevant’ dyads are looked at, for example, great powers and/or contiguous states (Maoz and Russett, 1991: tables 2 and 3). It is virtually impossible
without detailed historical case studies to determine aggressor and defender in
the roughly 30 wars since 1816 in which democracies have been involved
(Lake, 1992; Gleditsch and Hegre, 1995). It might not even make sense con-
20
Democratic Peace — Warlike Democracies?
ceptually to distinguish between challenger and target of aggression except for
very rare cases. Aggression and defence is usually in the eye of the beholder.
Look, for example, at the four most violent interstate wars since 1816 in which
democracies were involved. Only one of them — World War II — is a clear case
of aggression by a totalitarian regime. The Korean War (1950-53) is a case in
which the initial attack was undertaken by an authoritarian regime against
another non-democracy. A US-led coalition that included many democracies
then came to the defence of the target. However, this coalition then escalated
the war considerably by crossing the 38th parallel, which led to a — defensively
motivated? — countermove in terms of the war involvement of China.1 It is even
more difficult to identify aggressor and defender during World War I. As to the
Vietnam War, the U.S. not only intervened but also escalated it.
In order to establish that democracies are inherently more peaceful than
authoritarian regimes, one would need to show that democratic war involvement was mostly provoked by non-democracies. Rudolph Rummel appears to
be a dissenting vote from the scholarly consensus that democracies are as warprone as autocratic regimes (Rummel, 1995). He argues that democratic
regimes are less violent in their foreign behaviour than autocratic regimes. His
data suggest, however, that the main dividing line is not between democracies
and non-democracies, but among the latter, i.e., between authoritarian and
totalitarian regimes.2
In sum, there is not much evidence that democracies are less frequently
involved in military threats and the use of force in international relations or that
they are disproportionately more often on the defensive rather than the offensive side in those militarised disputes or wars. The war involvement of democracies has somewhat decreased since the 1970s; there also seems to be a correlation between the stability of democratic systems and their war involvement
(Chan 1984; Russett and Maoz 1993). But these negative correlations are not
robust enough to infer a general peacefulness of participatory political systems.
The Democratic Peace
While democracies do not seem to be inherently more peaceful than autocratic
regimes, there is nevertheless an island of peace in an ocean of conflicts and
wars. Democracies rarely fight each other, at least since 1816.3 There have been
military threats, the mobilisation of troops, and small skirmishes between
democracies, but (almost) no inter-state war. Moreover, such militarised disputes among democracies occur significantly less frequently than is to be
expected in a random distribution. And those militarised disputes that do occur
among democracies almost never escalate into war (Bremer, 1993; Maoz and
Abdolali, 1989; Maoz and Russett, 1991). The finding of a ‘democratic peace’ is
21
Thomas Risse
robust with regard to the time-period selected, the specific definition of democracy, or the method used (quantitative or qualitative). At least three different
data sets — the Correlates of War project, the International Crisis Behaviour
project, and the Hamburg-based Vergleichende Kriegsursachenforschung
project (Gantzel et al., 1986) — have been used to substantiate the claim (Bremer, 1992; Doyle, 1983; Gleditsch and Hegre, 1995; Maoz and Abdolali,
1989; Maoz and Russett, 1991; Nielebock, 1993). There is no other empirical
finding in the realm of international relations that has reached a similar consensus among scholars.
Following Kenneth Waltz’s earlier attack on the ‘democratic peace’ proposition, there have been two recent criticisms of the claim (Waltz, 1993; Layne,
1994; Spiro, 1994). Christopher Layne argues on the basis of four serious crises
among democratic states that these did not escalate for ‘realist’ rather than ‘liberal’ reasons. David Spiro claims that the ‘democratic peace’ finding is not statistically significant, given that wars occur rarely and that democracies are also
pretty rare in the international system. Both attacks miss the mark. First, two of
Layne’s four cases have been thoroughly investigated by John Owen who insists
that the lack of escalation in Anglo-American relations in 1861 (U.S. civil war)
and in 1895-96 (Venezuela crisis) had a lot to do with the mutual perception as
liberal democracies (Owen, 1993, 1994). As to the 1923 Ruhr crisis, another
case of Layne’s ‘near misses’, it is equally questionable whether the French public and elites perceived Germany during the 1923 Ruhr crisis as a liberal
democracy, given the instability of the Weimar Republic at the time. This leaves
the 1898 Fashoda crisis to support Layne’s argument. Not a very convincing
data base to challenge the ‘democratic peace’ proposition!4
Second, as to Spiro’s claims about the statistical insignificance of the ‘democratic peace’ finding, Bruce Russett’s rebuttal takes care of most points. If data
are split in ever-smaller parts, it is mathematically impossible to find statistical
significance. Russett then used Spiro’s method of year-by-year analysis with
regard to the ‘Militarised Interstate Dispute’ data and confirmed the ‘democratic peace’ proposition in a statistically significant way (Russett, 1995). In sum,
the challenge to the ‘democratic peace’ proposition rests on rather dubious
assumptions and equally doubtful empirical analysis.
Zeev Maoz and Bruce Russett have looked at alternative explanations for the
‘democratic peace’ hypothesis (Maoz and Russett, 1993). They show that
wealth and conflict behaviour are not correlated. While rich democracies are
slightly more involved in militarised disputes among each other than poor ones,
wealthy authoritarian regimes fight each other far more frequently. While geographic proximity slightly increases the rate of disputes among democracies,
this variable is far more significant with regard to autocratic regimes. Economic
interdependence also adds to peace, but this finding does not eliminate the
effect of democracy (Oneal et al., 1995). The only other variable that correlates
22
Democratic Peace — Warlike Democracies?
significantly with peacefulness is the stability of the regime measured in years.
Stable political systems, whether democracies or autocracies, are less likely to
fight each other. This finding implies that stable democratic systems constitute
the main inhabitants of the ‘island of democratic peace’.
In sum, democracies almost never fight each other, even though they are
sometimes involved in militarised disputes and the exchange of threats. They do
not automatically form ‘pacific federations’ (Kant, 1795[1991]) or ‘pluralistic
security communities’ (Deutsch, 1957). But there appears to be a barrier that
prevents militarised conflicts among democracies to escalate into war. As
William Dixon has shown, third-party mediation is very likely to succeed when
democratic systems threaten to use force against each other (Dixon, 1993,
1994). As a result, there is a contrast between the ‘democratic peace’, on the
one hand, and the war involvement of democracies with authoritarian systems,
on the other, that requires theoretical explanation.
Prevailing Explanations
A theoretical explanation of the empirical findings of a ‘democratic peace’
despite frequent war involvement by democracies must answer four questions:
1. Why is it that militarised disputes are far less likely to occur among democratic dyads than among democratic-authoritarian or autocratic-autocratic
dyads?
2. Why is it that militarised disputes among democracies almost never escalate
into war?
3. Why is it that democracies are more likely to co-operate in alliances and –
maybe - other international institutions than autocratic political systems?
4. Why is there less difference in the general war involvement of democracies
as compared to other types of political systems?
These four questions can be summarised in one: Why is it that the ‘security
dilemma’ appears to be far less significant when democracies deal with each
other, while it seems to govern their interactions with authoritarian systems?
The ‘security dilemma’ represents a good starting point for evaluating competing explanations of the empirical findings. It shows why and how even
peacefully motivated states in an anarchic self-help system can end up in arms
races, escalatory processes, crises and war with each other. To explain war, we
do not need to assume aggressive intentions or expansionist motives on either
side of the opponents (Herz, 1950; Jervis, 1978). It is not so much intentions
that drive the security dilemma, but lack of information and uncertainty about
23
Thomas Risse
the motives of other states which are then perceived as potential opponents. Of
course, arguments about the security dilemma are firmly grounded in realist
theory. Structural realism in the Waltzian tradition does not assume power-maximising states to explain war in the international system. Rather, it is sufficient to
conceptualise states as defensive positionalists concerned about their survival
in a self-help environment in order to explain arms races, crises, and war
(Waltz, 1979; Grieco, 1988a). A liberal response to the realist claims that concentrates on establishing the inherent peacefulness of liberal democracies is,
therefore, insufficient. Given the condition of uncertainty in an anarchic selfhelp system, states can be as peaceful as they want to be, they still have to be
defensive positionalists if they want to survive in such a system.5 Rather, a liberal response to realism must, first, explain why and how democracies ‘know’
that other participatory systems have equally peaceful intentions, i.e., why the
uncertainty that drives the security dilemma is far less relevant when democracies deal with each other. Second, it follows that such a response cannot exclusively focus on the unit-level of domestic polities and politics, but needs to
include the level of international interactions among states. I argue in the following that the prevailing liberal interpretations of the ‘democratic peace’ fail
on both accounts.
Participatory Constraints and Rational Cost-Benefit
Calculations
Immanuel Kant already argued that democracies do not fight each other
because of the participation of the citizens in the decision-making processes:
If ... the consent of the citizens is required to decide whether or not
war is to be declared, it is very natural that they will have great hesitation in embarking on so dangerous an enterprise. For this would
mean calling down on themselves all the miseries of war, such as
doing the fighting themselves, supplying the costs of the war from
their own resources, painfully making good the ensuing devastation... (Kant, 1795[1991: 100]).
The peaceful foreign policy of democratic states is explained by the rational
cost-benefit calculations of the citizens in a participatory polity. Aggressive foreign policy and the costs of war run counter to the welfare interests of the citizens — except for the immediate defence of one’s territory and way of life. It is
argued that the citizens in democratic states are primarily motivated to increase
their economic well being, as long as their external security is guaranteed. As a
result, they are unlikely to support military adventures and wars of aggression,
24
Democratic Peace — Warlike Democracies?
except, maybe, under circumstances of low risk (Müller and Risse-Kappen,
1993: 390-91). While this account focuses on the cost-benefit calculations of
the citizens, a ‘political incentive explanation’ of the democratic peace has
been suggested that concentrates on such calculations by national leaders
(Geva et al., 1993).6
Cost-benefit calculations and the price of war fighting might explain a certain reluctance of participatory polities to engage in military adventures. But
such calculations should apply irrespectively of whether democratic systems
deal with each other or with dictatorships. Rational choice arguments might
explain why democracies pursue aggressive foreign policies under specific
circumstances, but not why societal support for such interventions can only be
mobilised against authoritarian systems. Referring to utilitarian cost-benefit
calculations of citizens in a democratic polity cannot explain why foreign
aggression is absent even in extremely asymmetrical relations in which the
costs of attack are low for the more powerful state — as long as democracies are
involved on both sides. Moreover, the argument does not address the uncertainty issue that lies at the heart of the security dilemma. It merely accounts for
some reluctance to engage in warfare. As I will argue later in this article, costs
and benefits of aggressive foreign policies are not quasi-objective entities, but
depend on perceptions, norms, and collective identities. The latter define the
framework within which utility calculations take place.
Institutional Constraints and the Complexity of Democratic
Decision-Making
A second explanation emphasises institutional constraints. Democratic political systems are characterised by an elaborate set of checks and balances —
between the executive and the legislature, between the political system, interest groups, and public opinion etc. Decision-making processes, particularly
those involving war and peace issues, need time, since leaders need to mobilise
sufficient domestic support to go to war. The size of winning coalitions is
expected to be far larger in democracies than in authoritarian systems in which
support by small elite groups might suffice. One could then argue that the complexity of the decision-making process makes it unlikely that leaders readily use
military force unless they are confident of gathering enough domestic support
and of the costs being low. They will perceive leaders of other democracies as
equally constrained and, therefore, refrain from violence (Bueno de Mesquita
and Lalman, 1992: ch. 4; Morgan and Campbell, 1991; see also Russett, 1993:
38-40).
In a sense, the finding that democracies rarely fight each other would be a
statistical artefact according to the ‘institutional constraints’ argument, since
25
Thomas Risse
the democratic character of decision-making processes would not explain the
absence of war among liberal systems, but the higher complexity of their policy-making institutions. The more centralised the political systems, the more
likely it is that their leaders go to war. If so, the decisive distinction would not be
between democracies and non-democracies, but between highly centralised
and fragmented political structures, between ‘weak’ and ‘strong’ states as far as
the state-society relationships are concerned (Katzenstein, 1978; Krasner,
1978). The degree of autonomy which decision-makers enjoy vis-à-vis societal
demands should then correlate with their preparedness to use force in international politics. One would expect, for example, that the French state with its
highly centralised political institutions of the Fifth Republic would be more
likely to go to war than the comparatively fragmented and ‘society-dominated’
U.S. political system. There are not many empirical studies available investigating systematically the war involvement of ‘strong’ as compared to ‘weak’ states
(see, however, Russett and Maoz, 1993; Morgan and Campbell, 1991; Morgan
and Schwebach, 1992).
More important, the ‘institutional constraints’ model is unconvincing theoretically, since it tries to separate norms and institutions. Institutions consist of
norms of appropriate behaviour, rules defining role expectations, and stabilised
anticipations of the future. The complexity of democratic decision-making procedures is itself partly a function of democratic norms incorporated in the political institutions of liberal systems. The rule of law, for example, which is an
intrinsical characteristic of liberal democracies and is emphasised in Kant’s
‘Perpetual Peace’, is both normative and serves as an institutional constraint on
decision-makers. It is, therefore, almost impossible to distinguish between arguments focussing on the complexity of decision-making processes and those
emphasising norms.
Moreover, the model cannot explain why democratic leaders sometimes
quickly decide to go to war against authoritarian systems, are able to mobilise a
lot of societal resources within a short period of time — and get away with it by
counting on the ‘rally ‘round the flag’ effect (Mueller, 1973; Russett, 1990: ch.
2). The U.S. initial escalation of the Vietnam war represents such a case as well
as the interventions in Grenada and Panama. When liberal systems are faced
with authoritarian adversaries, the complexity of democratic institutions
appears to matter less. Finally, the ‘institutional constraints’ model might
explain why complex decision-making systems exert checks on national leaders concerning war fighting and, thus, emphasises — again — some inherent
peacefulness of such systems. The argument does not address the uncertainty
question discussed above, however. It is not clear why institutionally constrained leaders who perceive their fellow democrats as equally constrained,
should, therefore, refrain from violence. The model — in the absence of normative underpinnings — has no theoretically convincing argument why rational
26
Democratic Peace — Warlike Democracies?
leaders of democratic systems should not be tempted to exploit equally constrained leaders for a quick and less costly military victory, thereby overcoming
their own domestic restrictions. Rational leaders in institutionally constrained
decision-making structures could even fear that their equally constrained counterparts make similar calculations as a result of which uncertainty re-enters the
calculus. And back we are in the security dilemma!
The Externalisation of Democratic Norms
Liberal democracies are not only characterised by the rule of law, the constraints of checks and balances, and participatory rule of the citizens. Their
domestic structures also consist of norms, rules, and procedures embedded in
the political culture and institutionalised in the political system. Democratic
systems are supposed to guarantee freedom and human rights, to protect
minorities, and to establish the principal equality of citizens with regard to their
participatory rights. Democratic decision-making rules emphasise the peaceful
resolution of conflict through compromise and consensus, penalise the threat
or the use of force in domestic disputes as illegitimate, and provide for the possibility that today’s minority might become tomorrow’s majority. Democratic
governance emphasises social diversity, shifting coalitions, the consent of the
governed, and the publicity of the political process. Authoritarian regimes are
mostly characterised by the absence of these norms. If, however, the norms regulating the decision-making processes in democratic systems are oriented
toward non-violence and the peaceful resolution of political conflicts, one
could expect that democracies externalise these norms when dealing with each
other.
This argument emphasises the norms constituting the collective identity of
actors in a democratic polity instead of utilitarian cost-benefit calculations or
the complexity of decision-making processes. These norms are then expected to
shape the motivations, perceptions and practices of actors in liberal systems.
The argument offers a more convincing explanation for the ‘democratic peace’
than the other accounts, since it focuses on normative structures shaping interaction processes rather than individual behaviour. Norms which have been
internalised by actors to the extent that they affect their collective identities, are
expected to guide interactions, whether in the domestic or in the international
realm. Democratic systems are expected to externalise their internal decisionmaking norms and rules in their foreign policy behaviour. Two functions of
norms can be distinguished (Kratochwil, 1989). First, norms serve as communication devices that enable interactions in the first place by providing a framework of shared and collective understandings. Second, norms create, regulate,
and stabilise social order. It is the first function that might provide an explana-
27
Thomas Risse
tion for the externalisation of democratic norms when liberal polities deal with
each other. We need additional arguments, however, in order to explain why
democracies are also likely to externalise the regulatory functions of their
domestic norms when dealing with each other. It is not obvious that democracies are likely to create co-operative international institutions on the basis of
compromise-oriented decision-making norms — social order — among themselves. I come back to that point later.
The argument offers a convincing explanation why democratic state
actors are in general defensively motivated when it comes to international
relations. Norms as communication devices also provide a starting point to
explain why liberal democracies perceive each other as peacefully motivated. It is this perception that needs to be assumed in order to deal convincingly with the uncertainty factor inherent in the security dilemma. But the
normative explanation for the democratic peace is still under-theorised.
Bruce Russett, the leading advocate of the normative explanation, argues, for
example, that,
if people in a democracy perceive themselves as autonomous, selfgoverning people who share norms of live-and-let-live, they will
respect the rights of others to self-determination if those others are
also perceived as self-governing and hence not easily led into
aggressive foreign policies by a self-serving elite (Russett, 1993: 31;
see also Czempiel, 1981: 219; Doyle, 1986).
This argument rests on an assumption, which is not self-evident. In a dog-eatdog world of anarchy, the self-image as democratic and peaceful might or might
not guide interactions with other democracies. Again, the perception of others
as equally restrained might lead rational actors into adventurous foreign policies themselves or create the uncertainty that other rational leaders might come
to the same conclusion. One frequently used line of reasoning to overcome this
problem is to assume that aggressive behaviour is forced upon democracies by
the mere existence of authoritarian regimes. Autocratic leaders are not bound
by liberal norms of non-violent and compromise-oriented resolution of conflict,
since they do not form part of their identity:
In non-democracies, decision-makers use, and may expect their
opponents to use, violence and the threat of violence to resolve
conflict as part of their domestic political processes. ... Therefore
non-democracies may use violence and the threat of violence in
conflicts with other states, and other states may expect them to use
violence and the threat of violence in such conflicts (Russett, 1993:
35; see also Czempiel, 1992: 263).
28
Democratic Peace — Warlike Democracies?
In other words, authoritarian systems determine the rules of the game in their
interactions with democratic states. The latter have to keep up their guard in
order to contain potentially aggressive behaviour of the former. Aggressive
behaviour of peacefully minded democratic states would ultimately result from
the mere existence of autocratic regimes in the international system. As argued
above, such an explanation cannot be easily reconciled with the empirical
data. For example, one would expect to find clearer indications that militarised
disputes between democracies and authoritarian systems are more often
caused and initiated by the latter rather than the former. In striking contrast to
the robustness of the ‘democratic peace’ finding, there is not much in the empirical data to suggest that.
In sum, the normative explanation of the ‘democratic peace’ provides a
good starting point for theorising about war and democracy. The arguments presented so far might explain the defensive motivations of democratic systems
stemming from their collective identity as well as the potentially aggressive
intentions of authoritarian systems. But if the argument is right that the security
dilemma results from uncertainty rather than from intentions, these explanations are insufficient. There must be something in the interactions between
democracies and dictatorships, on the one hand, and among democracies, on
the other, that explains the difference in behaviour. Where do democracies
‘know’, on the one hand, that their fellow democracies are equally peaceful
and can, therefore, be trusted? On the other hand, why do they feel potentially
threatened by autocratic rulers thus enacting the ‘security dilemma’ when dealing with them?
A Social Constructivist Interpretation of the
Normative Argument
Social constructivism as such does not provide a substantive theory of international relations. It is not a ‘fair weather’ theory of international politics, as some
have misunderstood it.7 Rather, social constructivism provides a set of assumptions of what it is that we theorise about in international relations:
• International relations form part of the social construction of reality. Social
rather than material structures constitute actors as social individuals. Structures and agents are mutually constitutive (Wendt, 1987, 1992; Jepperson et
al., 1995).
It follows that anarchy and the resulting security dilemma are social constructs
themselves (see also Mercer 1995). The structure of the international system has
not somehow fallen from heaven, but has been created by the states and their
29
Thomas Risse
interactions. Consequently, anarchy and self-help as fundamental characteristics of international relations are not unavoidable.
• Actors’ interests and preferences cannot be treated as exogenous or fixed in
a theory of international relations. Rather, they originate and change during
the processes of social interaction. As a result, we must move beyond the
logic of instrumental rationality implied by rational choice theory and incorporate the logic of persuasion and communicative action. Words matter as
much as deeds (Müller, 1994).
• Ideas and norms have to be taken more seriously in international relations
than traditional theories based on material interests can account for.
In particular, the notion of ideas as ‘switchmen’ between material interests and
behaviour has to be challenged (Hall, 1993). A more fundamental notion holds
that ideas and norms have constitutive effects on the identity of actors and, as a
result, on the way actors perceive their interests (Wendt, 1994; Jepperson et al.,
1995). To put it differently, material interests do not exist in an idea-free vacuum.
What does a social constructivist interpretation of the democratic peace and
the war involvement of democracies add to the liberal argument, particularly its
normative version? Perceptions are the starting point. The proposition that
authoritarian regimes are potential aggressors given their domestic structure of
oppression and violence can be easier reconciled with the empirical data, if it is
not taken as a quasi-objective finding, but as a perception by democratic systems. As Michael Doyle put it,
[D]omestically just republics, which rest on consent, then presume
foreign republics also to be consensual, just, and therefore deserving of accommodation. ... At the same time, liberal states
assume that non-liberal states, which do not rest on free consent,
are not just. Because non-liberal governments are in a state of
aggression with their own people, their foreign relations become
for liberal governments deeply suspect. In short, fellow liberals
benefit from a presumption of amity; non-liberals suffer from a
presumption of enmity (Doyle, 1986: 1161).8
In other words, enmity as well as friendship in the international system does
neither result from some inherent features of the international distribution of
power, as realists would assume, nor from the domestic structures of states as
such, as the above-quoted liberals argue. Rather, it is socially constructed. The
democratic peace as well as the frequently aggressive behaviour by liberal
states toward non-democracies then results from a rule learned through the
processes of interaction, namely to infer aggressiveness or peacefulness from
30
Democratic Peace — Warlike Democracies?
the degree of violence inherent in the domestic political structure of one’s
opponent. This explanation has to be distinguished from the liberal interpretations discussed above in that it does neither assume some inherent aggressiveness of authoritarian systems nor some intrinsic peacefulness of liberal democracies. Rather, intersubjective perceptions count and the rule to infer
external behaviour from one’s internal political structures and its degree of violence and oppression. Peacefulness as well as enmity is learned through international interactions.
But why is it that the perception of others as either potentially hostile or
probably friendly depends on an evaluation of their domestic political structures? We get closer to an answer if we conceptualise norms as communication
devices. Decision-makers in democratic polities who have been socialised in
the norms governing liberal states, are likely to communicate their intentions in
the international realm by referring to these very norms. When they encounter
fellow democrats, a collective understanding of these norms can be easily
established providing a common basis for further communication of peaceful
intentions. Leaders of democratic states communicating their peaceful intentions to each other can always validate their claims by pointing to the peaceful resolution of conflicts inherent in their domestic structures. In other words,
the validity claims of peacefulness are substantiated by one’s own domestic
structure. As a result, the spiral model of the security dilemma is reversed and
uncertainty reduced. The possibility to validate one’s claim of peacefulness
through reference to one’s domestic norms also explains why the democratic
peace is particularly characteristic of stable democracies. States in various
stages of transition to democracy may still be met with suspicion given their
domestic turmoil.
The argument that democratic norms serve as communication devices by
which leaders of liberal states are likely to establish peaceful relations with fellow democracies can also account for the fact that there are indeed militarised
disputes among democracies, but that these almost never escalate into war.
First, nothing in the argument presented so far assumes that there are never real
conflicts of interests among liberal states. The issue is about how these conflicts
are dealt with, not their existence. Second, the argument about communication
processes by which democratic leaders establish the mutual recognition of
norms of peaceful conflict resolution implies that time is involved. The democratic peace results from a process of social interaction. Initially hostile confrontations particularly in cases of severe conflicts of interests are well within
the realm of possible behaviour. The longer the crisis endures, the more likely it
is that norms of peaceful resolution of conflict can finally be established or that
third-party mediation succeeds. Third, norms do not guide behaviour in such a
way that they are never violated. We do run red lights from time to time. Valid
norms guide behaviour in a probabilistic sense. As a result, the argument pre-
31
Thomas Risse
sented here does not require zero wars among democracies. It does require,
though, that leaders of democratic states use third-party mediation or engage in
specific communicative behaviour such as apologies, justifications, offer of
compensatory action when violating norms.
Insights derived from psychological attribution theory shed further light on
both the virtual absence of the security dilemma among democratic states and
its presence in interactions between the latter and dictatorships. Attribution theory posits that individuals tend to judge the behaviour of others on the basis of
dispositional rather than situational factors, while one’s own behaviour is
regarded as driven mostly by situational components (Crocker et al., 1983;
Jones et al., 1971; Kahnemann, 1986; Lebow and Stein, 1993; Stein, 1994).
Actors of democratic systems would then attribute potentially co-operative and
non-aggressive behaviour to fellow democrats in the international system, since
their domestic structures pre-disposes them toward such conduct. If the other
state is considered as democratic and just, one does not assume to be potentially threatened. As a result, one starts trusting each other. Trust, however, is a
functional equivalent of complete information and, thus, reduces uncertainty,
the driving force of the security dilemma (Luhmann, 1989). If actors of democratic states view each other as pre-disposed toward peacefulness, the significance of the security dilemma in their interactions is substantially reduced and,
therefore, a major obstacle toward stable security co-operation removed.
Actors who trust each other, start behaving accordingly. They thereby create a
peaceful and co-operative order through their interaction processes which reinforces the perception of one’s peaceful intentions. In other words, the presumption that the other is pre-disposed toward peacefulness leads to a self-fulfilling
prophecy, if both sides act on this assumption. The ‘democratic peace’ is socially constructed.
The democratic character of one’s domestic structures then leads to a collective identification process among actors of democratic states defining the ‘ingroup’ (Wendt, 1994). Such a collective identity not only substantially reduces
the significance of the security dilemma, it also decreases the importance of the
two major obstacles to international co-operation identified by regime theory fear of cheating and the ‘relative gains’ problem (Oye, 1986; Baldwin, 1993;
Grieco, 1988a. Mutual trust reduces fears of cheating. As to the ‘relative gains’
problem, actors of democratic states ‘know’ through the process of social identification that they are unlikely to fight each other in the future. The realist
assumption is no longer valid that today’s friend can always be tomorrow’s enemy and that, therefore, each state has to worry that its partner might gain more
from co-operation than oneself. Mutual gains through co-operative arrangements become possible.9
If we assume a collective identity because of shared liberal values among
democratic state actors, co-operative norms regulating the interactions among
32
Democratic Peace — Warlike Democracies?
democracies are likely to emerge when liberal systems are faced with co-operation problems. If these norms are institutionalised in regimes or international
organisations such as NATO or the European Union, democracies are expected
to form the Kantian ‘pacific federations’ (foedus pacificum) or what Karl W.
Deutsch called a ‘pluralistic security community’. Such communities are
defined as
a group of people which has become ‘integrated.’ By INTEGRATION
we mean the attainment, within a territory, of a ‘sense of community’ and of institutions and practices strong enough and widespread
enough to assure, for a ‘long’ time, dependable expectations of
‘peaceful change’ among its population (Deutsch et al., 1957: 5-6).
While Deutsch’s notion of pluralistic security communities was not confined to
democracies, it is unlikely that a similar collective identity and sense of mutual
responsiveness could emerge among autocratic leaders. There is nothing in
their values that would prescribe mutual sympathy, trust, and consideration.
Rather, co-operation among non-democracies is likely to emerge out of self-interests narrowly defined.
Why is it that domestic orders, norms, and political cultures shape the identities of actors in the international realm? Why not economic orders such as
capitalism? Why not geographic concepts such as ‘the West’, the ‘North
Atlantic area’, and the like? Why not gender and race such as ‘white males’? It
is, of course, trivial that actors hold multiple identities. Which of these or which
combination dominate their interests, perceptions, and behaviour in a given
area of social interaction, needs to be examined through empirical analysis and
cannot be decided beforehand. I submit, however, that values and norms pertaining to questions of governance are likely to shape identities in the realm of
the political - be it domestic or international. Moreover, notions such as ‘the
West’ do not contradict the argument here, but seem to represent a specific
enculturation of a broader liberal worldview. The same holds true for identities
as ‘capitalists’, particularly if juxtaposed against ‘Communist order’. The notion
of the ‘free world’ which Western policy-makers used frequently during the
Cold War to refer to their collective identity and to demarcate the boundaries
against ‘Communism’ encompassed liberal values pertaining to both the political and the economic orders (Latham, 1995).
Most liberal systems not only share democratic values, but also a common
appreciation of capitalist market economies.10 Why is it that their community of
values is based upon the norms governing their domestic political rather than
economic orders? If the latter were the case, relative gains would matter more,
since the norms governing capitalist orders emphasise competition and — albeit
regulated — conflict. Fierce economic competition among capitalist states does
33
Thomas Risse
not contradict the notion of security community among democracies, though.
International competition among capitalists is as much predicated upon a stable and peaceful international order as a domestic capitalist economy requires
the state to guarantee property rights and the rule of law. Such a stable international order cannot be built upon capitalist values of competition, but requires
the mutual recognition of democratic norms of peaceful conflict resolution.
If the pre-disposition of actors in democratic states leads them to trust other
democracies, to externalise their internal decision-making norms in their interactions among each other, and, thus, to construct the ‘democratic peace’, how
is the war involvement of liberal states with non-democracies to be explained?
Starting from the above-quoted argument by Michael Doyle, I argue that the
same rule — identity formation in international relations on the basis of one’s
domestic order — leads to the opposite predisposition and perception when
actors of democracies deal with authoritarian regimes. It is irrelevant in this
context whether dictatorships are inherently aggressive or not. It is enough to
set a security dilemma in motion, if one side perceives the other as potentially
violent. If democracies infer potential aggressiveness out of the internal political structure of oppressive regimes, they will feel threatened and act accordingly by forming alliances, engaging in arms races and even aggressive foreign policy. While ambiguous behaviour by other democracies is likely to be perceived
as confirming the assumption of friendliness, ambivalent behaviour of autocratic regimes will probably confirm the perception of their pre-disposition toward
violence and, hence, the feeling of being threatened. As the security dilemma
tells us, mutually aggressive behaviour might well occur despite mutually
defensive intentions. Belligerence then results from uncertainty about each other’s motives. Unlike realism which posits that the security dilemma is caused by
the eternal anarchic nature of the international system, a social constructivist
interpretation of the liberal argument points out that, first, the security dilemma
is confined to relations between democracies and autocracies or among the latter and, second, results from dispositional attributions in conjunction with
behavioural consequences creating the anarchical structure in the first place.
Enmity is socially constructed through mutual perceptions and interactions.
While democracies are likely to form a collective identity because of shared
values whereby other liberal states become part of ‘us’, dictatorships are perceived as the ‘other’, the out-group. This does not exclude tactical alliances
such as the US-UK alliance with the Soviet Union in 1941 or the American-Chinese relationship after 1972, but the co-operation problems identified by rationalist regime theory should apply. Since there is no mutual trust, fear of cheating might prevent co-operation. Since there is no sense of community, the
‘relative gains’ problem should play a more significant role. The Western
aggressiveness during the Cold War and its sometimes grotesquely exaggerated
threat perceptions can be easily accounted for in this context. The social con-
34
Democratic Peace — Warlike Democracies?
struction of the Communist threat also explains why détente and arms control
were so difficult to establish in the East-West relationship. If both sides perceive
their conflict as irreconcilable, as caused by dispositional rather than situational factors, and as involving values and fundamental goals rather than means,
conflict regulation becomes very difficult.11
An approach combining domestic structure characteristics, perceptions, and
interaction patterns in the international realm appears to offer a better explanation than conventional liberal accounts for the dual finding of the ‘democratic
peace’ despite frequent war involvement by democracies. However, if the presumption of friendship among democracies is as socially constructed as the
presumption of hostility in relations between the latter and autocratic systems,
one should assume that both did not automatically emerge with the existence
of democratic states in international relations. Rather, one would assume an
evolutionary process and learning (Levy, 1994; Russett, 1993: 5-11). Two factors might have encouraged that leaders in democratic states learned to perceive other democracies as peaceful and, by the same token, to distance themselves from autocracies.
First, publicity is an important ingredient of democratic decision-making
processes. When foreign policy becomes the subject of public controversies
(Holsti, 1992; Russett, 1990: ch. 4), and when aggressive foreign policies provoke counter-reactions in a democratic polity, actors can be easier persuaded of
the peaceful nature of democratic systems. The more decision-making processes become transparent, the more other political systems and societies can convince themselves that democracies are defensively motivated. The absence of
publicity in autocratic systems, however, increases the feeling of uncertainty by
liberal states and might lead to increased suspicions. Western speculations
about Iraq during the Gulf crisis and war as well as about North Korean ambitions with regard to nuclear weapons illustrate the point. Iraqi and North Korean behaviour was frequently evaluated on the basis of dispositional factors
whereby the lack of transparency confirmed perceptions of the aggressive
nature of these autocracies.
At the same time, the publicity of the democratic process is unlikely to be
perceived as reassuring by non-democracies, since transparency threatens the
rule of autocratic leaders. The CSCE process, for example, created international
publicity on the violation of human rights in Eastern Europe and the Soviet
Union thereby empowering dissident groups and undermining the rule structure of the Communist systems (Deudney and Ikenberry, 1991/92; Thomas,
forthcoming). In other words, the difference in publicity contributes to the democratic peace among liberal systems, on the one hand, but also increases threat
perceptions between democracies and authoritarian systems.
Second, liberal states generate transnational relations among societal actors,
since democracies are characterised by the separation between state and socie-
35
Thomas Risse
ty. The more autonomous societies are vis-à-vis the state, the less governments
are capable of controlling or prohibiting transnational activities of their citizens
(Keohane and Nye, 1971, 1977; Risse-Kappen, 1995b). This is not to argue that
transnational relations are restricted to democracies. I only posit that liberal
societies facilitate the emergence of transnational actors, while transnational
access to autocratic states tends to depend on the explicit consent of the rulers.
But transnational relations as such do not promote peace among states. They
might have different effects depending on whether they link democratic systems
with each other or democracies and dictatorships. Transnational relations within the democratic community of states might increase the sensitivity of the societies for each other thereby providing one more mechanism by which liberal
systems communicate peaceful intentions. Transnational contacts between
democratic societies and autocratic states, however, are unlikely to communicate defensive motivations, but, first of all, might undermine oppressive regimes
by decreasing the closed character of their societies. There are quite some indications, for example, that transnational exchanges between Western human
rights and peace movements, on the one hand, and dissident groups in Eastern
Europe, the Philippines, and South Africa, on the other, empowered the latter
and, thus, undermined the autocratic rule structures (Klotz, 1995; Sikkink,
1993; Thomas, forthcoming). As a result, transnational linkages between democratic societies and authoritarian states might not contribute to peaceful relations, but might even increase the tensions between the former and the latter,
since they challenge repressive rule from inside.
Conclusions: Suggestions for a Research Agenda on
Democracy and Peace
I have argued in this article that a social constructivist interpretation of the normative explanation for the dual finding of ‘democratic peace’ and ‘democratic
war involvement’ provides a better account than reference to cost-benefit calculations, institutional constraints, or an alleged inherent peacefulness of liberal systems. Democracies do not fight each other, because they perceive each
other as pre-disposed toward peacefulness and then act on this assumption.
They perceive each other as peaceful, because of the democratic norms governing their domestic decision-making processes. These norms constitute their
collective identity in international relations. They externalise them when dealing with each other, thus reinforcing the presumption of peacefulness. For the
same reason, they are likely to form pluralistic security communities based on a
collective identity. Because they perceive each other as peaceful and express a
sense of community, there are likely to overcome obstacles against internation-
36
Democratic Peace — Warlike Democracies?
al co-operation and to form international institutions. The norms regulating
interactions in such institutions are expected to reflect the shared democratic
values and to resemble the domestic decision-making norms.
At the same time, relations among democracies and authoritarian regimes
suffer from the presumption of potentially aggressive intentions. Democratic
state actors assume that autocratic leaders are pre-disposed toward belligerence,
since their domestic rule is based on oppression and violence. This perception
then creates a security dilemma leading to behavioural patterns that confirm the
presumption of enmity. Both the absence of the security dilemma among
democracies and its presence when democratic systems deal with dictatorships
are then socially created through perceptual and interaction processes. Thus, the
explanation is situated on both the domestic and the international levels.
The argument put forward in this article is highly speculative at the moment.
While it is consistent with the empirical data, the processes it describes require
detailed investigation. However, there are some empirical studies suggesting that
collective identities and the perceptions of enmity and friendship matter indeed.
John Owen (1993, 1994), for example, has shown in detail how perceptions of
the other as democratic or non-democratic mattered in war-threatening crises
involving the U.S. from late 18th century to World War I. His work comes close to
an empirical test of the argument developed above. Another study pertains to the
question of whether the Peloponnesian war involved wars among democracies,
particularly between Athens and Syracuse. Challenging the argument by Bruce
Russett and William Antholis (1993), Tobias Bachteler (1995) argues that Athens
and Syracuse did not perceive each other as stable democracies and, thus, could
not develop a collective identity preventing them from fighting each other. Internal turmoil and frequent domestic violence prohibited a mutual perception as
peaceful and liberal. My own work (Risse-Kappen, 1995a) on the European influence on U.S. foreign policy elaborates on the proposition that democracies are
more likely than other types of political systems to align with one another. I argue
that interaction patterns in the transatlantic alliance can be explained by assuming a security community in Karl W. Deutsch’s sense creating a collective identity
of shared democratic values which then leads to norms of consultation and consensus-based decision-making. I submit that this account provides a better explanation of the empirical evidence in the US-European Cold War relationship than
traditional alliance theories based on realist bargaining theory.
However, these studies only represent a first attempt at systematically evaluating the normative-constructivist explanation of the ‘democratic peace’ and
the ‘democratic war involvement’. We need detailed case studies on
• the domestic discourses leading to the war involvement of democracies as
compared to ‘near misses’ when disputes among democracies did not escalate;
37
Thomas Risse
• the origins of such militarised disputes among democracies as well as the
reasons for the lack of escalation to war (see Dixon, 1993, 1994);
• the role of publicity and of transnational relations in creating both the ‘democratic peace’ and perceptions of potential hostility between democracies
and autocratic systems;
• the differential impact of mass public opinion and elite opinion in the
processes described above.12
I particularly emphasise that such studies involve detailed analyses of
communication processes. One fundamental problem of most quantitative
studies on democracy and war is that they rely, first, on data on a highly aggregate level, and, second, on behavioural evidence. An explanation focussing on
norms, identity and perceptions, however, must involve discourse analysis,
since words matter in such an account as much as deeds. The main reason why
current studies evaluating the normative explanation of the democratic peace
against the ‘institutional constraints’ model have been found inconclusive is
that measuring norms via behaviour alone is questionable.13 Norms and identities affect behaviour via processes of communicative action. If we ignore these
processes in our research designs, we cannot convincingly prove or disprove
the normative argument.
Notes
1*
11.
12.
13.
14.
We wish to thank Sage Publications for granting us the permission to reprint this article, which
originally appeared in European Journal of International Relations, Vol. 1, (1995), No. 4, pp.
491-518.
It is now clear that the crossing of the 38th parallel by the Western democracies constituted the
major escalatory move during the Korean war leading to the Chinese intervention (Christensen,
1992).
Table 2, for example, which he considers superior to Table 1 which relies solely on Correlate of
War data, shows no difference between democracies and non-democracies for violence dead
as percentage of population (Rummel, 1995).
Two points have to made here. First, if general and universal suffrage is considered the most significant criterion for the democratic character of a political system, it should be noted that there
have been no democracies since approximately 1900. Until that time, women were mostly not
allowed to vote. How can one speak of democracies, if half the population is excluded from
participatory rights? It might well be that gender is relevant for the ‘democratic peace’ finding in
the sense that the women vote strengthens it (see Brandes, 1994; Russett, 1995: 167/fn.8). Second, most scholars argue that democracies ‘rarely’ fight each other, since there are some borderline cases depending on how strict criteria for what constitutes a democracy one uses.
According to James Lee Ray (1993), however, none of the alleged ‘wars among democracies’
qualifies, if peaceful change of government following free elections that involve at least 50% of
the population is used as the definitional criterion for ‘democracy.’
Layne’s treatment of pre-World War I Imperial Germany as a democracy as a result of which
World War I was a war among democracies is too grotesque to require a comment by a European scholar. For example, the controversy surrounding Fritz Fischer’s work that Layne calls
38
Democratic Peace — Warlike Democracies?
15.
16.
17.
18.
19.
10.
11.
12.
13.
‘ideologically tinged’ was not about the alleged democratic character of Wilhelmine Germany,
but about the domestic causes of World War I and their relative significance as compared to
other causes of war (Layne, 1994: 41-44).
At this point, I agree with Christopher Layne’s criticism of the prevailing liberal explanations for
the ‘democratic peace’ (Layne, 1994: 7-13; see also Owen, 1994: 90-103).
I thank an anonymous reviewer for alerting me to this article.
See, for example, John Mearsheimer’s recent analysis of ‘critical theory.’ He lumps together very
different approaches having allegedly in common the assertion that ideas move the world. See
Mearsheimer (1994/95).
For a similar point see Russett (1993: 33). As the above-quoted excerpt from Russett’s book
shows, however, he appears to argue closer from a position which assumes an intrinsic malevolence of at least some authoritarian systems rather than a socially constructed enmity between
democracies and dictatorships.
This is not to argue that only liberal democracies form stable international regimes. Rather, I
submit that democratic states can overcome obstacles to international co-operation more easily
than other types of systems.
I thank Mark Laffey for alerting me to this point.
On this point, see the ‘problem-structural’ approach to regime analysis as developed by the
Tübingen group (Efinger and Zürn, 1993).
In this article, I have deliberately avoided the issue whether the social construction of the democratic peace and the war involvement of democracies only concern the elite level or must
include the level of mass public opinion. I suspect the latter, but am unable to provide a convincing argument. I thank an anonymous reviewer for alerting me to this point.
Bruce Russett and Zeev Maoz, for example, operationalise the normative approach via the
persistence of a political regime and the level of violent internal conflict. Neither indicator can
be regarded as directly and causally linked to norms. In fact, the persistence of political regimes
over time can as well be explained within the ‘institutional constraints’ model (Russett and
Maoz, 1993).
39